05:30 vita_cell: imirkin_ karolherbst I tryed latest GNU/Linux Mint 17.2 and the result
05:30 vita_cell: no more MHZ core
06:13 karolherbst: vita_cell: what do you mean by that?
06:14 vita_cell: I still don't know why I can not to reclock core
06:14 vita_cell: tryed GNU/Linux mint 17.2 and other kernels
06:14 karolherbst: vita_cell: we know somehow
06:15 karolherbst: nothing you can do about that though
06:15 vita_cell: this is something of hardare problem?
06:15 karolherbst: no
06:15 karolherbst: driver
06:16 vita_cell: and can I update to some ohter version?
06:17 vita_cell: OpenGL core profile version string: 4.1 (Core Profile) Mesa 11.0.4
06:17 karolherbst: no
06:34 vita_cell: libdrm-nouveau2 2.4.65 I have installed
06:34 vita_cell: and xserver-xorg-video-nouveau 1:1.0.11
06:38 pmoreau: It will not help in getting reclocking to work, as reclocking is done in the kernel driver.
06:38 vita_cell: ok
06:39 pmoreau: What card do you have already? Was it a Kepler?
06:41 vita_cell: yes
06:41 vita_cell: gtx650
06:41 vita_cell: and it is not mine
06:41 vita_cell: I testing and benchmarked
06:41 RSpliet: pmoreau: iirc, fb reclocks, but the "core" doesn't
06:41 RSpliet: vita_cell: for the sake of brevity, we consider it yours :-P
06:41 vita_cell: yes hahhahaha
06:42 vita_cell: tomorrow I will test gtx770 asus
06:42 RSpliet: vita_cell: were you running karolherbst's kernel?
06:43 vita_cell: now, running Linux-Libre 4.3
06:43 RSpliet: ah hmm, that complicates matters slightly
06:43 vita_cell: but tryed with fresh install of GNU/Linux Mint 17.2
06:43 vita_cell: this is no blob problem
06:43 RSpliet: clock changing is under heavy maintenance
06:44 vita_cell: what it does mena
06:44 vita_cell: mean
06:44 pmoreau: RSpliet: Right. Didn't some of Karol's patches land in 4.3? Or will they only land in 4.4?
06:44 vita_cell: here some patched kernels repository?
06:45 vita_cell: [ 6885.672684] nouveau 0000:01:00.0: clk: failed to raise voltage: -22
06:45 vita_cell: [ 6885.672690] nouveau 0000:01:00.0: clk: error setting pstate 2: -22
06:45 vita_cell: [ 7368.976671] nouveau 0000:01:00.0: clk: failed to raise voltage: -22
06:45 vita_cell: [ 7368.976677] nouveau 0000:01:00.0: clk: error setting pstate 0: -22
06:45 RSpliet: 4.4 I'm afraid
06:45 RSpliet: oh there we go, yes, that issue hasn't been fixed even for kernel 4.4
06:46 RSpliet: (for which... even RC1 isn't out :-D)
06:46 vita_cell: so, it won't work for 4.3 (reclock core) ??
06:46 pmoreau: Looks like it won't
06:46 vita_cell: is here some repository for 4.4?
06:47 RSpliet: vita_cell: I'm afraid it isn't going to work just yet, someone has to reverse engineer the whole voltage situation properly before we can alter it to support the higher frequencies
06:48 RSpliet: I thought it was a WIP
06:48 RSpliet: but don't take my word on it, I only know old cards :-P
06:48 vita_cell: only Kepler supports reclocking?
06:49 pmoreau: RSpliet: That's a good thing! Really thankful to you for my G96 and MCP79. ;-)
06:49 RSpliet: no, as far as "reclocking" goes, the old G200 and GT21x are the best supported families
06:50 RSpliet: oh, and those MCP ones, those were relatively easy :-D
06:50 RSpliet: Keplers work well given DDR3 memory, the more higher end GDDR5 ones are more tricky
06:50 karolherbst: pmoreau: he has GPIO based voltaging
06:50 RSpliet: the GDDR5 ones are close to being supported, but no cigar... unfortunately it's an "all or nothing" problem :-)
06:50 vita_cell: Kepler with gddr5 worked for my
06:51 pmoreau: karolherbst: Ah right! I had forgotten that
06:51 RSpliet: pmoreau: hehe, glad I was able to help
06:51 pmoreau: RSpliet: How is the G94 doing by the way, apart from being noisy?
06:52 RSpliet: pmoreau: reclocks fine
06:52 RSpliet: 4.4 will support that one ;-)
06:52 pmoreau: Cool! :-)
06:52 RSpliet: unfortunately that means I ran out of cards... again :-D
06:53 vita_cell: something to do with gtx460?
06:53 pmoreau:is longing for 4.4, to make use of his G96…
06:53 vita_cell: I see that this one doesn't work more thar 99hz for me
06:54 RSpliet: vita_cell: no, that's one from the Fermi family
06:54 RSpliet: as far as changing clocks goes, those are the worst
06:54 vita_cell: I see that Fermi is bad supported for eberything
06:54 RSpliet: nobody has had much time to investigate them
06:55 vita_cell: GeForce GT240 (NVA3 (GT215) ) it is a good one?
06:55 karolherbst: not even close
06:55 karolherbst: :D
06:55 RSpliet: hmm... what kind of memory is on there?
06:56 karolherbst: ohh you talk about reclokcing support
06:56 karolherbst: mhhh
06:56 RSpliet: if GDDR3 it should be well supported since last year
06:56 vita_cell: yes gddr3
06:56 vita_cell: what the problem it is with gddr5?
06:56 RSpliet: it's more complex than the other memory types
06:57 RSpliet: more bits
06:57 RSpliet: and... I don't have a card with it, so I can't reverse engineer it :-D
06:57 vita_cell: ahhhh, but worked for me with my gtx650 gddr5
06:57 vita_cell: so gdrr3 has a better suppport
06:58 RSpliet: more specifically
06:58 vita_cell: gt520 it is Fermi? has bad support?
06:59 RSpliet: GDDR3 has better support for the Tesla cards (GT21x) than GDDR5
06:59 RSpliet: for Kepler, GDDR5 is supported better than GDDR3 :-P
06:59 vita_cell: ok, thaks, useful info
06:59 RSpliet: it's a confusing mess I realise, hence it's very difficult to communicate publicly
06:59 karolherbst: RSpliet: maybe the gddr5 apporach works also for gddr3
06:59 karolherbst: :D
07:00 karolherbst: but what is the issue with kepler gddr3?
07:00 RSpliet: karolherbst: on Kepler? unlikely, I expect it to be closer to DDR3
07:00 RSpliet: it's not implemented :-D
07:00 karolherbst: :D
07:00 karolherbst: I see
07:00 karolherbst: I have to say may card ran pretty stable today
07:01 karolherbst: let it run with +10% higher core clocks without crashes
07:01 vita_cell: what card it is yours?
07:01 vita_cell: 770m?
07:03 karolherbst: yes
07:07 vita_cell: and you can reclock core and memory of 770m?
07:07 karolherbst: yes
07:07 karolherbst: I am pretty sure your issue is something trivial
07:08 karolherbst: all voltage problems are, we just don't know how to parse the vbios the right way
08:49 yorick: so I tried nouveau, got a hang (at the end) http://sprunge.us/jRCF
08:55 karolherbst: yorick: you may want to try out a newer version of everything first
08:56 karolherbst: allthough, mhh
08:56 karolherbst: vita_cell: did you boot with nouveau.debug=volt=debug?
08:57 imirkin: yorick: channel unknown i think _might_ mean it's the kernel moving stuff around. and that failing is certainly not good. no idea why it'd happen though.
08:57 imirkin: oh, and ftr, CE = copy engine
08:58 vita_cell: yes, I did, but what it does? karolherbst
08:58 karolherbst: it will print a lot into dmesg
08:59 vita_cell: yes, but this parametr does nothing
08:59 vita_cell: I can see the same in dmesg
09:00 vita_cell: I did this
09:00 vita_cell: GRUB_CMDLINE_LINUX_DEFAULT="nouveau.pstate=1 nouveau.debug=volt=debug"
09:00 karolherbst: mhhh
09:01 karolherbst: vita_cell: then nouveau.debug=debug
09:01 vita_cell: I did something wrong?
09:01 vita_cell: ok
09:02 karolherbst: ohh wait
09:02 karolherbst: nouveau.debug=VOLT=debug helps
09:02 RSpliet: vita_cell: maybe a stupid question, but did you regenerate your grub.cfg afterwards?
09:02 vita_cell: ahahahha yes yes
09:02 karolherbst: mhhh
09:02 vita_cell: sudo update-grub
09:02 karolherbst: no, volt shold work
09:02 RSpliet: good good :-)
09:02 vita_cell: thanks you for asking
09:02 karolherbst: vita_cell: do you have a line like volt: current voltage: 881250uv
09:02 karolherbst: ?
09:02 vita_cell: not, where it is?
09:03 karolherbst: dmesg
09:03 vita_cell: [ 6885.672684] nouveau 0000:01:00.0: clk: failed to raise voltage: -22
09:03 vita_cell: [ 6885.672690] nouveau 0000:01:00.0: clk: error setting pstate 2: -22
09:03 vita_cell: [ 7368.976671] nouveau 0000:01:00.0: clk: failed to raise voltage: -22
09:03 vita_cell: [ 7368.976677] nouveau 0000:01:00.0: clk: error setting pstate 0: -22
09:03 vita_cell: this is only output errors
09:03 vita_cell: when trying reclock
09:03 karolherbst: no, you should get it at loading time
09:04 karolherbst: pretty much at the beginning
09:04 karolherbst: just give me your entier dmesg :D
09:04 karolherbst: and I will check
09:05 vita_cell: [ 1.669832] nouveau 0000:01:00.0: volt: current voltage unknown
09:05 vita_cell: [ 6885.672684] nouveau 0000:01:00.0: clk: failed to raise voltage: -22
09:05 vita_cell: [ 7368.976671] nouveau 0000:01:00.0: clk: failed to raise voltage: -22
09:05 karolherbst: ahhh
09:05 karolherbst: current voltage unknown
09:05 karolherbst: yeah, that might be a problem
09:06 vita_cell: and
09:06 vita_cell: [ 1.667698] nouveau 0000:01:00.0: volt: VID 00: 0uv
09:06 vita_cell: lines
09:06 karolherbst: yeah
09:06 karolherbst: okay
09:06 karolherbst: mupuf_: seems like nouveau can't handly his vbios correctly indeed
09:06 imirkin: vita_cell: this is why it's important to include full dmesg in a pastebin.
09:06 karolherbst: nvkm_volt_get returns an error
09:07 vita_cell: do you want me to upload full my output of dmesg?
09:07 karolherbst: okay
09:07 karolherbst: yeah
09:07 karolherbst: would help
09:07 imirkin: make sure to use pastebin, don't just paste it into irc
09:07 karolherbst: the VID thingies aren't parsed correctly for him, at least I know that much
09:08 vita_cell: but, the problem is, I cant see completely all outputs in Terminal
09:08 imirkin: vita_cell: dmesg > file. upload file to a pastebin (e.g. paste.debian.net)
09:08 vita_cell: how I do this command?
09:08 vita_cell: what to write?
09:10 vita_cell: Somethimes I have kernel's log files with 3gb in
09:14 karolherbst: dmesg > file
09:22 vita_cell: https://www.dropbox.com/s/o4j8bca0lg01g87/dmesgOutput.txt?dl=0
09:22 karolherbst: yeah, as I suspected
09:22 vita_cell: what?
09:23 vita_cell: unknown voltage?
09:25 karolherbst: mupuf_: there seems to be a problem inside nvkm_volt_parse_bios
09:26 karolherbst: ohh no, nvbios_volt_parse
09:30 karolherbst: anybody any idea what is the difference between the first and the second value? Vid 0, voltage 0 µV/837500 µV
09:30 imirkin_: karolherbst: rtfs :p
09:31 karolherbst: well it's not used by nouveau, the second value
09:36 karolherbst: I think I got it now
09:36 imirkin_: i meant nvbios
09:36 karolherbst: yeah, I know
09:37 karolherbst: but even there it doesn't tell me much
09:37 karolherbst: I just found out why the values are 0 with nouveau
09:43 karolherbst: mhh https://github.com/karolherbst/nouveau/commit/4046cbf6dbf54e3fa4c85ca430bf528a811a8feb
09:43 karolherbst: no idea if that works
09:45 karolherbst: vita_cell: you could try out to apply this patch on top of the nouveau source you are currently using
09:45 karolherbst: don't reclock
09:45 karolherbst: just boot with this debug options
09:45 karolherbst: and give me dmesg
09:46 vita_cell: I reboot with:
09:46 vita_cell: GRUB_CMDLINE_LINUX_DEFAULT="nouveau.pstate=1 nouveau.debug=VOLT=debug"
09:46 karolherbst: but you need to apply that patch and rebuild nouvau and stuff
09:46 vita_cell: where I get nouveau source?
09:47 karolherbst: you could use just use my branch though
09:47 karolherbst: git clone https://github.com/karolherbst/nouveau.git -b nve7_vita_cell_voltage
09:48 imirkin_: karolherbst: i think he's a few steps behind you... gonna have to do a lot more handholding :)
09:48 vita_cell: downloaded nouveau-master.zip
09:48 karolherbst: *sigh*
09:49 vita_cell: I only neet to compile, nothing to modify? right?
09:49 vita_cell: *need
09:49 karolherbst: do git clone https://github.com/karolherbst/nouveau.git -b nve7_vita_cell_voltage
09:49 vita_cell: en COnsole/terminal?
09:49 karolherbst: yes
09:50 vita_cell: can't find command
09:50 karolherbst: install git
09:50 vita_cell: some dependece
09:51 vita_cell: I must to install "dgit" package I think
09:51 vita_cell: or "git"
09:51 karolherbst: git
09:51 vita_cell: ok
09:52 vita_cell: vita@vita-desktop:~$ git clone https://github.com/karolherbst/nouveau.git -b nve7_vita_cell_voltage
09:52 vita_cell: Clonar en «nouveau»...
09:52 vita_cell: remote: Counting objects: 24757, done.
09:52 vita_cell: remote: Compressing objects: 100% (614/614), done.
09:52 vita_cell: remote: Total 24757 (delta 1351), reused 897 (delta 897), pack-reused 23246
09:52 vita_cell: Receiving objects: 100% (24757/24757), 5.20 MiB | 2.91 MiB/s, done.
09:52 vita_cell: Resolving deltas: 100% (19219/19219), done.
09:52 vita_cell: Checking connectivity... hecho.
09:52 karolherbst: nice
09:52 vita_cell: what to do now?
09:52 karolherbst: cd nouveau
09:52 karolherbst: cd drm
09:52 karolherbst: make
09:53 imirkin_: vita_cell: if you paste more than 3 lines into chan again, i'm going to ban you. i've warned you several times already about this.
09:53 vita_cell: nouveau folder in /home/USER/
09:53 vita_cell: ohhh sorry, how to do this correctly?
09:54 imirkin_: if you need to provide more than 3 lines, use pastebin
09:54 vita_cell: it is some option of Xchat?
09:54 imirkin_: http://hastebin.com
09:54 imirkin_: paste in there
09:54 imirkin_: and provide a link
09:55 vita_cell: ahhhh ok thanks and sorry
09:57 karolherbst: vita_cell: I already told you, after that git command, you should do:
09:57 vita_cell: make doesn't work
09:57 karolherbst: ohh
09:57 karolherbst: what does it print?
09:58 vita_cell: http://hastebin.com/ebexabovug.sm
09:58 vita_cell: no file or directory error
09:58 karolherbst: vita_cell: you need to isntall the dev package of your kernel
09:58 vita_cell: ahhh
09:58 karolherbst: or source package
09:58 karolherbst: don't know how it is called for yours
10:00 vita_cell: I serching
10:00 vita_cell: searching
10:08 vita_cell: I will use 4.1 kernel
10:09 karolherbst: vita_cell: you need to use the package of your current kernel
10:09 karolherbst: and it has to be 4.3
10:09 vita_cell: yes, I will install 4.1 and use 4.1 source
10:11 karolherbst: just use 4.3, because this branch won't work for older kernels
10:16 karolherbst: vinceh: use 4.3, because this branch won't work for older kernels
10:16 vita_cell: http://hastebin.com/usecohosoh.sm
10:20 vita_cell: karolherbst
10:21 vita_cell: error while compile nouveau
10:21 karolherbst: yeah and I told you to use 4.3
10:21 vita_cell: some repository for 4.3?
10:22 karolherbst: you did installed your 4.3 kernel somehow, there should also be a source package for it or anything
10:22 vita_cell: my repository it is for linux-libre, and I see no source for 4.3
10:23 karolherbst: I can't help you if your distributions does something oddly, if there is no source or developer package or whatever, it's a bug in the distribution
11:20 mupuf_: karolherbst: hey, I am here
11:20 mupuf_: and I have time tonight
11:25 karolherbst: mupuf_: yeah nice
11:25 karolherbst: I came up with something
11:26 karolherbst: mupuf_: https://github.com/karolherbst/nouveau/commit/4046cbf6dbf54e3fa4c85ca430bf528a811a8feb
11:26 karolherbst: I am not sure about that though
11:26 mupuf_: well, you should indeed be :D
11:26 karolherbst: nvbios_volt_entry_parse is only called when the header contains only 0
11:26 karolherbst: well or if there is no base or step or something
11:26 karolherbst: thing is
11:27 karolherbst: version 0x40 and 0x50 aren't handled in that case yet
11:27 karolherbst: vita_cell: got 4.3 kernel isntalled with sources to compile nouveau?
11:28 mupuf_: karolherbst: what are you trying to achieve?
11:29 karolherbst: to read the voltage out for the VID
11:29 karolherbst: -- Vid 0, voltage 0 µV/837500 µV (unk = 2288) --
11:29 mupuf_: do you have evidence that the voltage is read differently between the versions 0x40 and 0x50?
11:29 karolherbst: I only do that for 0x50 now
11:30 karolherbst: mupuf_: check that https://www.dropbox.com/s/o4j8bca0lg01g87/dmesgOutput.txt?dl=0
11:30 karolherbst: and search for the VID entries
11:30 vita_cell: http://hastebin.com/oweserifij.vbs
11:30 karolherbst: they are all id : 0 and 0uv
11:30 vita_cell: something I doing wrong
11:31 karolherbst: vita_cell: what distribution are you using?
11:32 karolherbst: mupuf_: and then I wanted nouveau to parse the vbios and use the second value nvbios prints: -- Vid 0, voltage 0 µV/837500 µV (unk = 2288) --
11:32 vita_cell: Trisquel 7 64bits with linux-libre repository
11:33 karolherbst: vita_cell: is there a linux-libre64-headers package?
11:33 vita_cell: yes
11:33 karolherbst: did you install it=
11:33 karolherbst: ?
11:34 vita_cell: yes
11:34 mupuf_: karolherbst: ack
11:34 mupuf_: let's have a look at this :)
11:34 karolherbst: :)
11:34 karolherbst: I've upladed his vbios
11:34 karolherbst: it's a nve7
11:34 vita_cell: linux-libre64, linux-firmware-image-4.3.0-gnu, linux-image-4.3.0-gnu
11:34 vita_cell: not, I can't see header
11:35 karolherbst: yeah you should install the headers one
11:35 vita_cell: header=source?
11:35 xexaxo: curious if any of the freedom (deblobbing) has made it into the headers ?
11:35 karolherbst: vita_cell: something like that
11:35 vita_cell: not, I only downloaded source manually
11:36 karolherbst: just install that linux-libre64-headers package
11:39 Wolf480pl: karolherbst, IIRC the inital value was 38
11:39 Wolf480pl: but I can check
11:39 karolherbst: Wolf480pl: mhhh
11:40 karolherbst: 44-38 = c
11:40 mupuf_: karolherbst: sorry, I really misread your patch
11:40 karolherbst: 3d - 31 = c
11:40 karolherbst: well
11:40 mupuf_: looking more into it
11:40 karolherbst: mupuf_: for both our cards, nouveau uses a 0xc lower pwm value
11:40 mupuf_: it makes a lot more sense than what I previously understood :D
11:40 karolherbst: :D
11:41 karolherbst: I just hope the values are right
11:41 mupuf_: how does that translate to voltage?
11:41 karolherbst: mupuf_: unrelated
11:41 karolherbst: ohh
11:41 karolherbst: that you mean
11:41 karolherbst: sry
11:41 karolherbst: mhhh
11:41 karolherbst: around 0.07V
11:41 mupuf_: and what is the the space between min and max? constant too?
11:42 karolherbst: mhh for my card the blob used 5 values depending on the temperature
11:42 karolherbst: with lower bounds
11:43 vita_cell: I have headers installed
11:43 karolherbst: wait....
11:43 karolherbst: I think
11:43 karolherbst: mupuf_: I finally managed to understand why I always have these 14MHz core freq steps
11:44 karolherbst: mupuf_: PCLOCK.CLK0_COEF => { M = 0x1f | N = 0x48 | P = 0x1 } PCLOCK.CLK0_COEF => { M = 0x1f | N = 0x47 | P = 0x1 } PCLOCK.CLK0_COEF => { M = 0x1f | N = 0x46 | P = 0x1 }
11:44 karolherbst: and so on
11:44 mupuf_: :)
11:44 karolherbst: I nearly never saw something else then P=1 and M=1f
11:45 karolherbst: and if N decreases, the PWM value is also decreased by one
11:45 karolherbst: but that's is more of a theory yet
11:45 karolherbst: I might dig more into that
11:45 karolherbst: vita_cell: try make again
11:45 vita_cell: same error
11:46 vita_cell: I put source 4.3 in /lib/modules/4.3.0-gnu/build
11:46 karolherbst: ohhh
11:46 karolherbst: you should _never_ mess with the package manager
11:47 vita_cell: so, what it does mean?
11:47 karolherbst: you shouldn't have put the source there yourself
11:47 vita_cell: to where I must to put it?
11:47 karolherbst: nowhere
11:47 karolherbst: you just install packages through your package manager
11:48 vita_cell: I have installed headers, but the same errors
11:48 vita_cell: http://hastebin.com/ixadewequx.sm
11:49 vita_cell: it searches /build folder
11:50 karolherbst: output of "ls /lib/modules/"
11:51 vita_cell: vita@vita-desktop:~$ ls /lib/modules/
11:51 vita_cell: 4.3.0-gnu
11:52 vita_cell: uname -r is 4.3.0-gnu too
11:56 karolherbst: vita_cell: remove that build directory and reinstall the linux-headers package according to your kernel
11:56 vita_cell: ok
11:58 karolherbst: vita_cell: linux-libre64-4.3-headers
12:05 vita_cell: building
12:06 vita_cell: compiling
12:07 vita_cell: karolherbst
12:07 vita_cell: http://hastebin.com/cawasusobe.avrasm
12:13 karolherbst: nice
12:13 karolherbst: vita_cell: you have a nouveau.ko file somewhere in /lib/modules/4.3.0-gnu
12:13 karolherbst: I think
12:13 karolherbst: kernel/drivers/gpu/drm
12:14 karolherbst: there is a nouveau folder
12:14 karolherbst: and what is inside that?
12:14 vita_cell: /lib/modules/4.3.0-gnu/kernel/drivers/gpu/drm/ ???
12:15 vita_cell: /lib/modules/4.3.0-gnu/kernel/drivers/gpu/drm/nouveau/nouveau.ko
12:15 karolherbst: mk
12:15 karolherbst: ok
12:15 vita_cell: it is?
12:15 karolherbst: rename nouveau.ko to nouveau.ko.old
12:15 karolherbst: and move the compiled nouveau.ko file where the old one was
12:15 karolherbst: /home/vita/nouveau/drm/nouveau/nouveau.ko
12:15 Wolf480pl: (isn't it supposed to be installed in extramodules?)
12:16 karolherbst: Wolf480pl: I don't want to mess with that
12:16 Wolf480pl: oh, your way doesn't require changing depmod, ok
12:17 karolherbst: no
12:17 karolherbst: just regenerating initramfs
12:17 vita_cell: only nouveau.ko? to move
12:18 vita_cell: ok I moved the new one
12:18 pmoreau: You should rather make a link, it would be easier
12:18 karolherbst: vita_cell: ls -la /lib/modules/4.3.0-gnu/kernel/drivers/gpu/drm/nouveau/
12:19 karolherbst: pmoreau: mhh it will be overwritten on a kernel upgrade anyway
12:19 karolherbst: or going "lost"
12:19 vita_cell: http://hastebin.com/efiqowuvod.avrasm
12:19 karolherbst: looks good
12:19 karolherbst: vita_cell: there should be a command to regenerated initramfs
12:19 karolherbst: but
12:19 karolherbst: I don't know how it is called
12:19 pmoreau: karolherbst: True, but if you gave him multiple patches to test before he upgrades, he will have to copy each time
12:20 pmoreau: s/gave/give
12:20 karolherbst: vita_cell: try run update-initramfs
12:20 karolherbst: pmoreau: yeah, but I am sure I will only give him this one, maybe a second one at most
12:20 vita_cell: and where it is the command to regenerated initramfs, where to search?
12:22 karolherbst: vita_cell: update-initramfs
12:22 vita_cell: http://hastebin.com/guqozoluxo.rb
12:26 gryffus: vita_cell: "update-initramfs -u"
12:28 vita_cell: update-initramfs: Generating /boot/initrd.img-4.3.0-gnu
12:28 vita_cell: W: mdadm: /etc/mdadm/mdadm.conf defines no arrays.
12:28 vita_cell: done?
12:41 m3n3chm0:nasZ
12:42 gryffus: vita_cell: now you should reboot and post new dmesg
12:42 vita_cell: ok
12:45 vita_cell: rebooted
12:47 vita_cell: karolherbst
12:47 vita_cell: karolherbst: you are a GENIUS !
12:55 karolherbst: vita_cell: ohh it works?
12:55 vita_cell: AC: core 540 MHz memory 1620 MHz
12:55 vita_cell: I can reclock to the second level
12:55 vita_cell: but not to 3
12:56 vita_cell: core stucks at 540mhz
12:57 karolherbst: okay
12:57 karolherbst: then try 0a pstate and check if it's stable
12:57 vita_cell: I think yes at 5000mhz of memory and 5020
12:58 vita_cell: but not at 5040
12:58 vita_cell: I tested with Stunt Rally 2.5 game
13:00 vita_cell: when switching from 07 to 0f, the core clock get stuck at 324mhz
13:01 vita_cell: when switch from 07 to 02 it reclocks fine
13:01 vita_cell: karolherbst
13:07 karolherbst: vita_cell: please upload your dmesg somewhere again
13:08 vita_cell: ok
13:08 vita_cell: full?
13:08 vita_cell: or only latest lines?
13:09 karolherbst: full
13:10 vita_cell: https://www.dropbox.com/s/1huwij9y2q1t73f/dmesgNouveau.txt?dl=0
13:12 vita_cell: from 07 to 0a it reclocks with 540mhz, works fine, but it don't wat to reclock from 07 to 0f, and I can not get 1071mhz
13:12 karolherbst: except
13:13 karolherbst: mhhh
13:13 karolherbst: there is something fishy though
13:13 vita_cell: what it does mean?
13:18 karolherbst: there is more to it
13:18 vita_cell: karolherbst
13:18 vita_cell: http://hastebin.com/ivurulocan.mel
13:23 karolherbst: vita_cell: try some opengl load under 0a
13:23 karolherbst: and check if it's stable
13:23 karolherbst: for a longer time
13:23 vita_cell: how long?
13:23 vita_cell: I am on AC: core 540 MHz memory 5000 MHz right now
13:58 vita_cell: karolherbst
14:21 karolherbst: vita_cell: okay, I got a fix
14:21 vita_cell: wow
14:21 vita_cell: I tested AC: core 540 MHz memory 5000 MHz playing 40min, stunt rally 2.5, Nexuiz
14:22 imirkin_: "tested"
14:22 vita_cell: yes, it is not sull stress but...
14:22 vita_cell: *full
14:22 karolherbst: vita_cell:
14:22 karolherbst: okay
14:23 karolherbst: go into nouveau/drm again
14:23 vita_cell: I can run glmak2
14:23 karolherbst: vita_cell: first we enable 0f for ya
14:23 karolherbst: :D
14:23 vita_cell: I don't need to download nothing again?
14:23 karolherbst: imirkin_: with that a lot of kepler voltage tables should make _much_ more sense: https://github.com/karolherbst/envytools/commit/d0bbefaf912098221902ab5517ed726b9a8841ff
14:23 karolherbst: vita_cell: wait
14:23 karolherbst: are you there?
14:24 vita_cell: what?
14:24 karolherbst: go into the nouveau repository into the drm folder
14:24 karolherbst: where you did make
14:24 karolherbst: we need to update your tree a bit
14:24 vita_cell: ok
14:24 imirkin_: karolherbst: lol
14:25 karolherbst: imirkin_: I know :D
14:25 vita_cell: vita@vita-desktop:~/nouveau/drm$
14:26 karolherbst: vita_cell: git fetch origin
14:26 vita_cell: git https://github.com/karolherbst/envytools/commit/d0bbefaf912098221902ab5517ed726b9a8841ff
14:26 vita_cell: ??
14:27 imirkin_: vita_cell: 'git fetch origin'
14:28 vita_cell: git fetch origin https://github.com/karolherbst/envytools/commit/d0bbefaf912098221902ab5517ed726b9a8841ff
14:28 imirkin_: vita_cell: no, just 'git fetch origin'
14:28 vita_cell: ahhh sorry, now understood
14:29 vita_cell: I did
14:29 vita_cell: done
14:29 imirkin_: karolherbst: you're on
14:30 karolherbst: okay
14:30 gryffus: imirkin_: BTW what about the patch you sent me for the fence crash? Will it be fixed in next Mesa? http://hastebin.com/tirikofini.coffeeThanks for info :)
14:30 karolherbst: git reset --hard origin/nve7_vita_cell_voltage
14:30 imirkin_: gryffus: i need to push a bunch of fixes for nouveau... haven't gotten around to it yet
14:31 gryffus: imirkin_: ok, i was just asking if you didn't forgot for this one :) Thanks
14:31 vita_cell: HEAD is now at a23cdce bios/volt: handle voltage table version 0x50 with 0ed header
14:31 imirkin_: gryffus: not forgotten, just... no time
14:31 gryffus: imirkin_: i can udnerstand that :)
14:32 vita_cell: 07: core 324 MHz memory 648 MHz I running right now
14:32 vita_cell: must enable 0f first?
14:33 karolherbst: vita_cell: make
14:34 imirkin_: karolherbst: probably "cd drm" first?
14:34 karolherbst: imirkin_: nah, I am smart I already told him to go into drm
14:35 vita_cell: I did git fetch origin and git reset --hard origin/nve7_vita_cell_voltage before enabling 0f
14:35 karolherbst: vita_cell: you have to recompile nouveau
14:35 karolherbst: install it again
14:35 karolherbst: update initramfs
14:35 karolherbst: and reboot
14:35 vita_cell: make > update-initramfs -u
14:36 karolherbst: then you need to copy the nouveau.ko file again inside /lib/modules
14:36 karolherbst: what we done before
14:37 vita_cell: ok
14:37 karolherbst: the updater the copy run update-initramfs
14:38 vita_cell: after copying nouveu.ko I run update-initramfs
14:40 imirkin_: skeggsb: should i keep sending warc800 patches, or are you just going to drop the big one?
14:40 karolherbst: vita_cell: okay, reboot
14:40 vita_cell: /home/vita/nouveau/drm/nouveau/nouveau.ko > to?
14:41 skeggsb: imirkin_: i guess for 4.3, we keep to what we're doing now (so: patches welcome), and risk default for 4.4
14:41 skeggsb: i'm open to other suggestions though
14:42 imirkin_: skeggsb: fine by me... but e.g. some of these are GK107's so won't help for 4.3
14:42 imirkin_: skeggsb: unless you also backport the pgob-for-gk107 patch
14:42 skeggsb: oh, 4.3->4.4, 4.4->4.5 ;)\
14:42 skeggsb: i'm lagging :P
14:42 imirkin_: ah!
14:43 imirkin_: skeggsb: otoh if you can write these yourself... even faster :)
14:43 imirkin_: skeggsb: https://bugs.freedesktop.org/show_bug.cgi?id=70354#c88
14:45 karolherbst: vita_cell: your dmesg please
14:45 vita_cell: first I try to run 0a and later 0f?
14:45 karolherbst: nah, first dmesg
14:45 vita_cell: ok
14:45 karolherbst: I want to check if you forgot something
14:47 vita_cell: https://www.dropbox.com/s/p86fk0tvkdziw89/dmesg1.txt?dl=0
14:47 karolherbst: vita_cell: looks nice
14:47 karolherbst: 0f might work now
14:47 karolherbst: current voltage also makes sense now
14:47 karolherbst: nouveau 0000:01:00.0: volt: current voltage: 1075000uv
14:48 vita_cell: 0a working
14:48 vita_cell: wait
14:48 vita_cell: your fix working
14:48 vita_cell: 0f works
14:48 vita_cell: AC: core 1071 MHz memory 5000 MHz
14:49 karolherbst: nice
14:49 karolherbst: but
14:49 karolherbst: your gpu might crash
14:49 karolherbst: because the voltage might be too low
14:49 vita_cell: yes I know, better low that high
14:49 karolherbst: I would do some benchmarks and check how stable that is
14:49 vita_cell: glmark2
14:49 karolherbst: gputest has pixmark piano which is really good for putting a lot of stress on the core
14:50 karolherbst: don't know what glmark2 does
14:50 vita_cell: so, now you can share this fixes to next drivers
14:51 karolherbst: yeah, somehow
14:51 karolherbst: skeggsb: wanna pull some voltage fix for kepler into 4.4? :D
14:51 karolherbst: skeggsb: https://github.com/karolherbst/nouveau/commit/a23cdce8aef69a95851f200ebd414d4a01a20e84
14:52 karolherbst: will investigate some vbios before I mail them though
14:52 vita_cell: karolherbst do you have some repository?
14:53 karolherbst: okay, have to sleep now :D
14:53 karolherbst: good night
14:53 vita_cell: thaks you very much!
14:53 vita_cell: *thanks
15:21 vita_cell: imirkin_
15:39 gryffus: OMG Why Fermi overclocking has to be so complicated, even on windows? :/ Just reading http://forums.guru3d.com/showthread.php?t=336117
15:43 gryffus: s/over/re
16:35 pmoreau: imirkin_: How do you access $r0l or $r0h using BuilUtil / nv50_ir commands?
16:35 imirkin_: pmoreau: split
16:36 imirkin_: pmoreau: mkSplit([r0l, r0h], r0)
16:36 imirkin_: or something
16:36 pmoreau: Oh, I thought it was to split a double
16:37 pmoreau: "OP_SPLIT, // $r0d -> { $r0, $r1 } ($r0d and $r0/$r1 will be coalesced)"
16:37 imirkin_: sure
16:37 imirkin_: the dirty little secret about nv50 is that it's all 16-bit registers
16:38 pmoreau: Eh eh! Ok
16:39 pmoreau: Thanks for the info
19:44 imirkin: gryffus: http://cgit.freedesktop.org/mesa/mesa/commit/?id=bb73fc4cb82c1abdf47aa373c78c2a85fe29b3ec
19:44 imirkin: gryffus: that should be the better fix
20:18 sarnex_: imirkin_: i can consistently cause a lockup by setting prime to the nouveau card and dragging the glxgears window back and forth
20:29 imirkin: sarnex_: don't do that ;)
20:30 sarnex_: i mean is there a bug somewhere or is that expected
20:30 imirkin: sarnex_: do you have any logs or something? is this a nouveau issue or perhaps radeon helps along?
20:30 sarnex_: ill ssh in with another system and look at dmesg i guess
20:31 imirkin: sarnex_: system lockups are never "expected"
20:31 sarnex_: yeah i wasnt sure if you were kidding lol
20:32 imirkin: yes and no... if you find somethign that hangs your desktop... stop doing that :)
20:32 sarnex_: well im interested in helping debug so i do dumb stuff
20:32 imirkin: hangs are really annoying to pin down, easiest to just avoid them
20:33 imirkin: plenty of issues elsewhere
20:41 sarnex: imirkin_: http://pastebin.com/raw.php?i=ZM9MtTrS
20:44 imirkin: sarnex: note a distinct lack of "nouveau" in those traces
20:47 sarnex_: imirkin_: if i do the same thing with my r600 card and prime it doesnt die
20:47 sarnex_: is it still radeon's bad?
20:48 imirkin: the messages are radeon
20:49 sarnex_: seems pretty obscure i doubt anyone from amd had time
20:49 sarnex_: thanks
20:49 imirkin: but you believe that the nouveau team is considerably larger?
20:50 sarnex_: no
22:09 Cymage: ok, here is the deal. i have a clevo/sager with the GTX 950M. i dont game, so i dont need nvidia driver. it is apparently installed by default in sabayon. would be ok if it wasnt stuck at 1024x768 instead of 1920x1080. so my question is, how do i dump nvidia and load the nouveau without breaking the install again?
22:10 imirkin: Cymage: look for a guide tailored for your distro...
22:11 Cymage: imirkin : figured it was going to come down to that. all of the guides i am finding are going the other way
22:11 imirkin: Cymage: btw if there's an intel chip in there (most likely), that'll work way better all around
22:11 Cymage: there is, and i would LOVE to just use it, if i could figure out how
22:11 imirkin: pastebin 'lspci' output
22:14 Cymage: imirkin : https://pastebin.sabayon.org/pastie/17547
22:15 Cymage: imirkin : wait, that wont help
22:15 Cymage: imirkin : wrong box
22:15 imirkin: Cymage: errr... you *only* have an intel chip in there
22:17 Cymage: imirkin : https://pastebin.sabayon.org/pastie/17548
22:17 Cymage: imirkin : thats from the bad box
22:17 imirkin: oh fancy. skylake.
22:18 imirkin: aka skybreak based on earlier irc discussion :)
22:18 imirkin: so actually i don't see how you can *not* be using your intel chip. the nvidia chip is marked '3d accelerator' which generally suggests no outputs are actually hooked up to it
22:20 Cymage: i have only been fighting that thing since ups dropped it off at noon. its only 1:30 in the morning now
22:20 Cymage: lol
22:20 Cymage: any suggestions? i am lost
22:21 imirkin: Cymage: pastebin dmesg + xorg log
22:21 Cymage: ok
22:22 Cymage: imirkin : dmesg == https://pastebin.sabayon.org/pastie/17549
22:23 imirkin: [ 1.577499] [drm] This hardware requires preliminary hardware support.
22:23 imirkin: See CONFIG_DRM_I915_PRELIMINARY_HW_SUPPORT, and/or modparam preliminary_hw_support
22:23 imirkin: you want a newer kernel
22:23 imirkin: grab a copy of 4.3.0
22:24 imirkin: also i'd recommend blacklisting the nvidia module (add the line "blacklist nvidia" to /etc/modprobe.conf
22:24 imirkin: make sure you have xf86-video-intel installed
22:24 Cymage: imirkin : ok, here is the xorg log https://pastebin.sabayon.org/pastie/17550
22:25 imirkin: right, so the core issue is you need to upgrade to 4.3.0
22:25 imirkin: because you have skylake, and skylake is only fully supported starting 4.3.0
22:26 Cymage: nice. that means i have to build it. doesnt look like it is in repo
22:28 imirkin: you could also try booting with i915.preliminary_hw_support=1 but... don't expect things to work
22:31 Cymage: at this point i will try anything
22:37 Cymage: firmware load error i915
22:38 Cymage: but, i have the proper resolution, just stuck in terminal
22:42 imirkin: anyways, your issues have nothing to do with the nvidia gpu... good luck :)
22:43 Cymage: thank you
23:05 imirkin: mupuf_: nvkm_wr32(device, 0x20344, 0x8000000 | duty);
23:05 imirkin: that's not off-by-one is it? i.e. that shouldn't be 0x80000000 perchance?
23:13 imirkin: skeggsb: that's in gk104_volt_set btw
23:13 imirkin: and according to what's recorded in rnndb, i'm right
23:19 skeggsb: that seems suspicious indeed
23:20 imirkin: gnurou: please confirm that this isn't a typo in ctxgk20a.c: nvkm_mask(device, 0x5044b0, 0x8000000, 0x8000000);
23:20 imirkin: gnurou: i.e. note that you're setting bit 27, not bit 31.
23:23 imirkin: skeggsb: i assume these are ok?
23:23 imirkin: drm/nouveau/nvkm/subdev/devinit/nv05.c: (!fbmem_readback(fb, 0x1000000, ++patt) ||
23:23 imirkin: drm/nouveau/nvkm/subdev/devinit/nv20.c: for (off = amount; off > 0x2000000; off -= 0x2000000)
23:24 imirkin: i'm looking at git grep -P '0x[0-9a-fA-F]{7}\b' :)
23:28 gnurou: imirkin: looks correct - bit 31 is not used in this register (it is a debug register)
23:28 gnurou: imirkin: what makes you think this is incorrect?
23:29 gnurou: ah - just saw your regex :P
23:29 imirkin: gnurou: normally nouveau would use 0x0800000 there. mind if i change it?
23:30 gnurou: you mean 0x08000000? (yours had one less 0) if so, sure, it's more legible
23:30 imirkin: gnurou: yes, that's what i meant :)
23:31 imirkin: otherwise it's easy to visually think you're setting the high bit when you're not
23:32 gnurou: until the day we switch to proper macros, Ken has a new version: https://github.com/Gnurou/nouveau/pull/3
23:32 gnurou: imirkin: yeah, definitely agree here
23:35 imirkin: skeggsb: cleanup patch sent... left the gk104 volt one alone coz i think that one's an actual bug. mupuf_ can hopefully confirm.
23:36 imirkin: gnurou: iirc the problem with that was his headers only document a tiny fraction of everything
23:37 mupuf_: imirkin: what the heck
23:37 mupuf_: that definitely looks like it
23:38 imirkin: mupuf_: happy to send a patch, but afaik you guys had tested this, so... would probalby be good to retest :)
23:38 mupuf_: a PWM controller was using 0x1000000 as the commit bit, but this is not what I remembered
23:38 mupuf_: yes
23:38 mupuf_: give me a sec
23:38 mupuf_: very good way to wake up :D
23:38 imirkin: haha
23:39 imirkin: speaking of which... bed time for me. i trust you can take care of sending the one-line fix patch if necessary.
23:40 mupuf_: ok, will do the one-liner if necessary
23:40 mupuf_: thanks for noticing this brown bag moment
23:40 mupuf_: I did test it, so ... what the heck?
23:43 mupuf_: I guess I just checked if the value changed but did not check with nvawatch if it changed anything at the GPIO level
23:46 hakzsam: imirkin, nice patch, very good catch :)
23:54 mupuf_: ok, confirmed that it really is supposed to be 8 + 7 zeros
23:55 mupuf_: this time, using the gpios and not just checking what the register says you wrote into it